The neighbor effect has long been invoked to explain species coexistence and biodiversity maintenance in communities (Peters, 2003; Comita and Hubbell, 2009; Zhu et al., 2015; Fichtner et al., 2018; Glatthorn, 2021). Heterogeneous distributions of trees mainly arise due to the interactions between neighbors and focal trees (Silander and Pacala, 1985; Kim and Underwood, 2015; Pu et al., 2020). Resource enrichment and habitat amelioration can enhance positive interactions (Callaway and Walker, 1997; Brooker et al., 2008; Maestre et al., 2009; le Roux et al., 2013), while the spread of natural enemies and resource competition for limited resources can cause negative interactions (de Souza and Válio, 2001; Hyatt et al., 2003; Getzin et al., 2006; Gunton and Kunin, 2007; Fricke et al., 2014). These interactions regulate tree mortality patterns that maintain species coexistence (Zhang et al., 2009; Paine et al., 2012).

The conspecific neighbor effect is an important index of intraspecific interaction (Comita et al., 2014; Ding et al., 2019). Negative intraspecific interactions occur among aggregated conspecific individuals through resource competition and/or through encountering natural enemies (Janzen, 1970; Connell, 1971; Zhu et al., 2015; Liu et al., 2016, 2021). The strength of these negative intraspecific interactions increases with the density of conspecific neighbors. Trees tend to have a higher probability of mortality when conspecific neighbors are denser, closer, or more abundant (Castagneri et al., 2010; Comita et al., 2010; Johnson et al., 2012; Wang et al., 2012). This situation thereby provides space for heterospecific recruitment and promotes the coexistence of different species (Yao et al., 2020). However, positive intraspecific interactions may also occur among conspecifics due to similar habitat preferences and/or beneficial microorganisms (such as nitrogen-fixing bacteria) (Wu et al., 2016). Therefore, positive and negative interactions may exist simultaneously and act significantly as part of the conspecific neighbor effect.

The heterospecific neighbor effect is a synthesis of interspecific interactions. Previous studies have found that negative interspecific interactions typically result from strong heterospecific competition for resources (Tilman, 1987; Hubbell et al., 2001; Getzin et al., 2006). In addition, positive interspecific interactions are attributed to the herd protection hypothesis, which posits that heterospecific neighbors can effectively hinder the spread and reduce the risk of detection and of invasion of specific natural enemies such as pests or pathogens, thereby reducing individual mortality (Wills, 1996; Peters, 2003; Zhu et al., 2015; Yao et al., 2020). Further, if heterospecific neighbors exhibit similar habitat preferences as individuals, positive interspecific interactions may take place within those habitats (Lebrija-Trejos et al., 2014; Wu et al., 2016; Yao et al., 2020). Finally, some studies have found that the strength of negative and positive interspecific interactions likely correlates with the density of heterospecific neighbors (Peters, 2003; Wu et al., 2017).

Neighbors affect tree mortality via various mechanisms associating with their life stages. For example, conspecific neighbors of earlier life stages are typically offspring of focal adult trees. In this situation, the patterns of conspecific, earlier life stage neighbors around the focal adult tree are most likely due to dispersal limitation (Vincent et al., 2011). Here, the conspecific earlier life stage neighbors hardly affect the mortality of the parent tree (Weiner, 1990). In contrast, conspecific neighbors of the same or later life stages compete with the focal individual for resources and can further the spread of species-specific natural enemies, sometimes resulting in tree mortality (Weiner, 1990). Conspecific crowding may also be a result of shared habitat preference, and in that case, tree mortality decreases due to increased habitat suitability (Lebrija-Trejos et al., 2014; Wu et al., 2016; Yao et al., 2020). Similarly, the crowding of later-stage heterospecific neighbors leads to shading and fewer encounters between a host and its species-specific pests and pathogens. These heterospecific effects often reduce tree mortality (Wills, 1996; Peters, 2003).

Previous studies have shown that effects of conspecific and heterospecific neighbors influence tree mortality. However, the relative effects of neighbors of different life stages have yet to be described. To fill this research gap, we examined the effect of neighbors of different life stages on tree mortality in a subtropical forest through a generalized linear mixed modeling framework. Additionally, as environmental filtering is also an important mechanism contributing to tree mortality (Wang et al., 2012; Baldeck et al., 2013; Shen et al., 2014; Wu et al., 2017; Yao et al., 2020), environmental variables were considered in the models to account for environmental effects on tree mortality. Specifically, we addressed the following questions: (1) Do conspecific and heterospecific neighbors affect tree mortality? (2) If so, are there differences among the effects of neighbors at various life stages, both of the neighbors and of the focal trees, and how do these differences interact?

Conspecific and heterospecific neighbors significantly affected ontogenetic tree mortality (Zhu et al., 2015). However, neighbors of different life stages different than that of the focal tree showed inconsistent effects (Koch et al., 2016). In our study, we found that conspecific earlier and later stage neighbors had opposite effects on juvenile mortality. Heterospecific same and later stage neighbors also had opposing effects on sapling mortality. The later stage neighbors were the only group whose singular effect was mostly consistent with the effect of all neighbors combined. In addition, failing to consider some life stages of neighbors may have resulted in the observation of related environmental factors showing more consequential effects.

Mortality attributed to conspecific negative density dependence is significantly widespread in the sapling stage, but this effect weakens as the tree grows (Zhu et al., 2015; Yao et al., 2020). Meanwhile, heterospecific neighbors reduce tree mortality throughout all life stages. Previous studies have shown that mortality due to conspecific negative density dependence is caused by specialized natural enemies and/or intraspecific competition for limited resources (Janzen, 1970; Connell, 1971; Zhu et al., 2015). Saplings, as the earliest life stage, suffer to a greater degree from natural enemy damage and from intraspecific competition than do older trees (Weiner, 1990). Therefore, saplings are more strongly impacted by conspecific neighbors. This result is consistent with findings of previous studies showing that density-dependent survival at the seedling and sapling stages plays a significant role in fostering tree species coexistence and maintaining diversity in forests (Bai et al., 2012; Johnson et al., 2014; Lin et al., 2014; Yan et al., 2015; Yao et al., 2020). Unlike conspecific neighbors, denser heterospecific neighbors can effectively hinder the spread of specific natural enemies and reduce objective tree mortality (Wills, 1996; Peters, 2003). This herd protection has also been observed in other studies of conspecific negative density dependence (CNDD) (Comita and Engelbrecht, 2009; Zhu et al., 2015; Yao et al., 2020). Furthermore, heterospecific neighbors displayed a negative correlation with adult tree mortality. This result indicates that heterospecific trees survive better around adult trees, which consequently promotes species coexistence.

Clumping of conspecific later stage and heterospecific same stage neighbors contributes the most to sapling mortality. Saplings are the youngest trees in this study, and as such are more susceptible to natural enemies and are less competitive. In this way, saplings may be subject to invasion upon the crowding of conspecific later stage neighbors because the neighbors could spread specific natural enemies, while also having asymmetric advantages in the competition for resources (Weiner, 1990; Clark and Clark, 1992; Uriarte et al., 2004; Yao et al., 2020). Consequently, conspecific later stage neighbors are the major contributor of all conspecific neighborhood effects to sapling mortality. In addition, conspecific later stage neighbors comprise the largest component of conspecific neighbors (Supplementary Figure 2). Thus, conspecific neighbors cause sapling mortality altogether (Figure 3A), although this effect can be largely attributed to conspecific later stage neighbors. This may also be why most studies of CNDD have found that the clumping of conspecific neighbors causes sapling mortality (Bai et al., 2012; Lin et al., 2014; Yan et al., 2015; Yao et al., 2020), even though these studies did not distinguish among conspecific neighbors’ life stages. In contrast, clumping of heterospecific later stage neighbors may reduce sapling mortality by hindering the spread of specific natural enemies (Peters, 2003). However, heterospecific sapling neighbors demonstrate the opposite effect, increasing focal sapling mortality. For one, herd protection is weakened in same stage heterospecific neighbors (Ramage et al., 2017), and they also have symmetric advantages in the competition for light resources, which can result in greater sapling mortality (Comita and Engelbrecht, 2009; Rüger et al., 2009; Bai et al., 2012; Piao et al., 2013).

Conspecific earlier and later stage neighbors show opposing effects on juvenile mortality, and this conflict destabilizes the overall conspecific neighbor effect. Conspecific earlier stage neighbors for juveniles are saplings, which are typically smaller and weaker than juveniles. Therefore, they hardly influence juvenile mortality (Weiner, 1990). However, the observation that conspecific sapling neighbors are clumped demonstrates that this habitat is suitable for regeneration (Grubb, 1977; Silvertown, 2004; Pérez-Ramos et al., 2012). The mortality of juveniles decreases correspondingly, and consequently, conspecific sapling neighbors significantly and negatively affect juvenile mortality. In contrast, adults are at a later stage than juveniles, and as mentioned above, they could also spread specific natural enemies and they have asymmetric competitive advantages compared to juveniles. However, juveniles are more adept at resisting natural enemies and asymmetric competition than are saplings (Boege and Marquis, 2005; DeMalach et al., 2016; Yao et al., 2020). Therefore, the positive effect of conspecific adult neighbors on juvenile mortality is weaker than it is on saplings. Furthermore, due to the weaker positive effect of conspecific adult neighbors and the counteraction of conspecific sapling neighbors, the overall conspecific neighbor effect on juvenile mortality is offset by these opposing effects. In our study, we were unable to even detect an overall conspecific neighbor effect on juvenile mortality (Figure 3B). Again, overlooking the life stages of neighbors when measuring their effect on focal tree mortality can result in a biased, incomplete conclusion.

Due to the positive spatial association between adult trees and their offspring, more living adult trees occur in the area with aggregations of earlier stage conspecific neighbors (Perea et al., 2021). Here, we found that when more conspecific earlier stage neighbors were clumped around the focal adult tree, its mortality was lower. Mature trees are generally surrounded by offspring due to dispersal limitation (Vincent et al., 2011). These dispersal patterns can have profound implications for later period, positive intraspecific spatial association (Moles et al., 2004; Beckman and Rogers, 2013; Sanín et al., 2013). Meanwhile, this positive intraspecific spatial association was observed to exhibit a greater degree around living adult trees than dead ones (Hou et al., 2004). Thus, clumping of conspecific earlier stage neighbors may reflect original dispersal limitation of focal adult trees with high survivability. In addition, although we considered the effects of environmental filtering on tree mortality by including several environmental factors in the analysis, other environmental factors that could have contributed to clumping of earlier stage conspecific neighbors and adult tree survival, such as canopy light, were excluded (Dechnik-Vázquez et al., 2016; Liu et al., 2017).

Environmental filtering is a vital process contributing to tree mortality across complex topographies (Wang et al., 2012; Yao et al., 2020; Andrus et al., 2021). The topography of the Tiantong plot is dynamic (Yang et al., 2011; Fang et al., 2017), with two parallel valleys in the north-south direction (Figure 1), and as such, convexity is considered to be an important environmental variable (Fang et al., 2017). In our study, trees tended to have a higher probability of mortality in low convexity habitats across all life stages. This may be because trees in lower convexity habitats may be disturbed more easily and experience less light availability. As light conditions are generally reduced from a south to north slope, juveniles occurring toward a north slope suffer higher mortality than juveniles near a south slope. In addition, initial tree size is consistently a strong predictor for tree survival (Wang et al., 2012; Ma et al., 2014; Wu et al., 2017). We also detected this effect across all life stages.

Environmental variables might have indirectly affected tree mortality through their correlations with the distributions of large neighbors. In this study, we classified the neighbors into different ontogenetic life stages. We found that removing some later stage neighbors from the analysis allowed the effects of some environment variables to become significant. These environmental variables were found to correlate with the variables of these later stage neighbors (Supplementary Figure 3). Therefore, this correlation supports the finding that the distribution of large individuals is influenced by environmental variables (Wang et al., 2012; Yao et al., 2020), meaning environmental variables likely affected tree mortality indirectly through neighborhood effects.

The findings of this study suggest that both conspecific and heterospecific neighbors play a major role in tree mortality (Peters, 2003; Zhu et al., 2015). However, these influences differ based on the life stage of the neighbors. Conspecific earlier stage neighbors generally have the effect of decreasing focal tree mortality because their clumping either reflects that the original dispersal limitation of the focal tree with high survivability or that the micro-habitat suitability for population colonization. In contrast, conspecific, later stage neighbors exert the effect of increasing focal tree mortality through CNDD. Furthermore, these opposing effects could minimize the overall conspecific neighbor effect. Due to herd protection, the assembling of heterospecific later stage neighbors could reduce focal tree mortality. In addition, large tree neighbors always dominate the neighborhood effects. At the same time, some environmental variables are closely correlated with the distribution of large trees. This correlation could result in the neighborhood effect on tree mortality, partially including an indirect environmental effect. Overall, these results suggest that there are multiple mechanisms among neighbors of different life stages in ontogenetic tree mortality, and thereby they highlight the necessity for dividing neighbors into different life stages when assessing the overall neighborhood effect contributing to species coexistence.

The datasets presented in this study can be found in online repository. The name of the repository and accession number can be found below: https://github.com/kkkliuheming/tree-mortality.

HL, YH, and XW conceived and designed the study. QSY, YC, HY, ZZ, and HL collected the data. YH and HL provided analysis tools and analyzed the data. HL, YH, QWY, QSY, and XW drafted and revised the article. All authors agreed to be accountable for all aspects of the work.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.